The Chinese government is between a rock and a hard place.  Recent mandates concerning the rationalization of power in certain provinces have put a hold on production in factories to reduce energy use as the country used up energy quotas faster than expected thus far this year, and to present a strong face at the UN Environmental Conference being held in Kunming, China today and for the upcoming Winter Olympics in Beijing.  Chinese utility companies, who are limited in their ability to pass on increased costs to consumers, are also being pressured by rising coal and gas costs, which the government has allowed in order to discourage excessive production. 
While the result from a physical standpoint seems to be clearer skies, implementing such harsh measures at a time when the Chinese economy is so dependent on the exportation of goods for the holidays seems a bit counterintuitive, and while it seems semiconductors, displays, and LEDs have all been spared such restrictions to a large degree, the effects of such power restrictions on upstream components will eventually take a toll on downstream production by further limiting product assembly and production.  The US press has taken hold of that possibility with stories about how you might not be able to get everything you want for the holidays because of said problems, and, of course, the inability of the global transportation system to operate with a smaller employee footprint.
That said, things can get worse.  The province of Shanxi in China, the country’s top coal producing region (394.8m tons thru May, ~26.4% share of total) is normally seeing the end of the rainy season, which typically peaks in July and is over in October.  However this year (global warming?) the province is facing unusually heavy rainfall (3x the monthly average in 5 days) which has caused crop damage, over 120,000 people being evacuated, the destruction of over 17,000 buildings, and the closing of at least 60 coal mines, with more rain predicted this week. 
While the government has taken radical steps to give the impression that it puts the environment ahead of economic growth, the timing could not be worse and Mother Nature is not helping.  This will only serve to exacerbate and extend product shortages that are already plaguing the global supply chain, all to maintain a face that few actually believe.  We can get along with a few less gifts over the holidays but those in China who depend on those factories being open will suffer the most.  While a single party dictatorship has the advantage of being able to implement what it wants when it wants and does not have the downside of two ineffectual parties that bicker over every detail, the downside is that stupid mistakes and inefficiency take a toll on much of the population. What’s the weather like in Reykjavik?

As we have noted in the past, Taiwan companies are required to report sales results monthly.  There are two large panel LCD producers in Taiwan, AU Optronics (AUOTY) and Innolux (3481.TT) and one small panel producer, Hannstar Display (6116.TT), all three of which we have been tracking for many years.  Given the complexity of the current consumer electronics cycle, while we are most concerned with quarterly results from the panel industry, we are even more focused on monthly trends.  We do gather monthly data on all panel producers, but that data becomes available after the month is over, while the Taiwanese data is reported only a few days after the month ends, giving a fresher look at results, albeit limited to a small number of companies.
Both large panel producers, AU Optronics and Innolux have seen strong sales increases through the June/July period this year, and while results are still positive on a y/y basis, the momentum has slowed in 3Q.  AUO saw both a q/q and y/y increase in sales in 3Q while Innolux saw essentially flat q/q results but positive y/y comparisons.  We expect 4th quarter y/y comparisons to still be positive, but q/q might be more difficult as panel price increases for IT products have stalled and panel prices for TV panels have declined.  October is typically a flat month for panel industry sales (-0.6% 5 year average) and the 4th quarter is typically down less than 1%, but we would expect a bit more volatility this year.  While Hannstar is essentially small panel production only, the site is currently down so we will have to update full September and 3Q performance at a later date.

On Friday we noted that LG Innotek (011070.KS) was committing just under $1b to expand its flip-chip ball grid array business, after noting that Samsung EM (009150.KS) had also committed to expanding it FC-BGA capacity, possibly in connection with Intel (INTC).  While both are able to garner significant financing from parent organizations to make such expansions, smaller FC-BGA suppliers are also increasing capacity as the technology become more widely adopted.  South Korean supplier Daeduck Electronics (353200.KS) has indicated that they will spend a bit over $250m US by next year to expand capacity after completing the construction of a new $75m plant in August.  The company has completed financing for the next line, which is expected to be in operation by the end of this year, and by the end of 2023 will have increased it capacity by 4x.  Korea Circuit (007810.KS) has also committed to a $167.2m expansion program after signing a long-term contract with a global communications chip manufacturer.
As we noted last week, during the shortages that the semiconductor industry has been facing, there is little opposition to funding advanced packing solution expansion projects such as these, and as complexity increases, low yields limit unit volumes to a degree which justifies expansion.  While growth over the next five years is pegged at  between 6% and 8%, FG-BGA products are typically higher margin for OSATs compared to typical consumer applications, with demand from AI, HPC, networking and automotive the drivers for the space, making such projects even more attractive.  Intel’s EMIB (Embedded Multi-Die Interconnect Bridge) and other 2.5 and 3D stacking methods are encouraging high-speed direct interconnects internal to the package, while eliminating TSVs and interposers, keeping packages smaller and connections faster.  FC-BGA compliments such advanced processes, which gives some upside to the market, but expansion seems to be happening at an unprecedented rate over the last few months, giving rise to a careful watch over the balance between supply and demand in 2022.  The diagram and table in Fig. 2 shows the advantage of using FC-BGA without TSV interposer and underfill in an image sensor product.

​Many Samsung (005930.KS) smartphones did not use Samsung’s own chipsets, including high-end flagship models.  The top of the line Samsung Galaxy S21 5G models all contained a Qualcomm (QCOM) Snapdragon 888 chipset if it was sold in the US and China, while other regions received the same phone with Samsung’s own Exynos 2100 chipset, which would seem rather unusual, but has been the case for many of Samsung’s smartphones.  In fact, much of Samsung’s smartphone line, including foldable use application processors produced by Qualcomm or Mediatek (2454.TT), and only Chinese brand Vivo (pvt) has released a few models with the Samsung Exynos chipset.  It seems odd that Samsung’s self-produced chipset would not appear in many of its smartphones, as it assumedly would be less expensive than purchasing an outside AP, but there have been problems that have led to a low adoption rate for the in-house AP, and Samsung aims to change that going forward.
Things got so bad that a petition on Change.org in April of last year requested that Samsung stop using the Exynos processor on many international versions and consistently use the Qualcomm processor, as “The Exynos phones are slower, have worse battery life, worse camera sensors (a separate issue) and processing, get hotter and throttle faster, amongst other issues.”  Comparison tests on sites like AnTuTu and GeekBench confirmed the weaknesses when compared (Exynos 990) to the Qualcomm 865 and Samsung’s share of the AP market has been declining and our quick check of the Exynos’s use in Samsung’s own phones seems to show a big push in 2019 and then a decline in its use in 2020 and this year.
One issue that has been cited for the problems associated with the Exynos AP is that Samsung had been using its own custom CPU cores (Mongoose) produced at its Texas facility, which as a single core outperformed Qualcomm, but when combined with other functions in the SoC, the package fell short in comparisons.  Samsung closed its custom CPU lines at the end of 2019 and adopted ARM based CPU design, similar to that used by Qualcomm and others.  Samsung has also been working with AMD (AMD) to improve graphics performance and will use AMD’s next generation graphics processing unit in the upcoming Galaxy S22 release expected in January.  Samsung is also said to be working on a fix for the heat problem, which is caused by 5G communication chips.
The S22 is still expected to be released with both the Exynos and Qualcomm options, but Samsung is expected to increase the use of the next generation Exynos processor in a larger portion of its smartphone line, which is also expected to have a unit volume target that is ~20% higher than this year.  While it is still early for setting 2022 targets, we expect some of the expansion of the Exynos AP is due to the shortages facing the semiconductor industry which could limit the availability of APs to Samsung.  By using its own AP, it can bypass some of the issues that could arise in 2022, but the comparisons between the ‘new’ Exynos and both Qualcomm and Mediatek must be much closer if Samsung does not want to face continued criticism.  By switching to ARM cores, they have narrowed the gap, but if they are going to expand the Exynos across the line, it better be as good as or better than the others.

Epitaxial[1] growth is a process by which silicon and III-IV compound devices are produced, with silicon substrates the leading material.  Much has been said about the necessity for high quality wafers on which various devices can be deposited but the underlying demand is what is expected to drive the expansion of the epi-wafer industry over the next few years.  Based on a model developed by Yole, demand in 2019 was for ~7.8m 6” equivalents, driven primarily by power electronics applications, but that will both grow and change by 2025 to ~21.3m 6” equivalents, driven by photonics, traditional LEDs, and Mini/Micro LED demand.


[1] Epitaxy is the deposition of materials on a crystalline substrate.

Epitaxy equipment, ~$692m last year is expected to grow to $1.1b by 2026, roughly an 8% CAGR but there are only a few suppliers of this equipment and over 60% of the market is controlled by the top three suppliers, Aixtron (AIXA.GR), Veeco (VECO), and AMEC (688012.CH).  As the technology for each application develops over the next few years, more specialized equipment could become necessary, which would give other tool vendors a better entry point when competing in this market.  Aixtron’s dominance in the Chinese market has allowed them to grow share by 10% over the 2018 – 2020 period, while both others have lost share, with a portion of those gains and losses a result of the trade war between the US and China. 
As noted smaller players are able to gain share with equipment such as MBE (Molecular Beam Epitaxy), which will be used for some forms of micro-LED production, but for traditional LED and Mini-LED, which use the same basic deposition process, the leaders have yet to be challenged, even by such equipment powerhouses such as Applied Materials (AMAT) and Tokyo Electron (8035.JP), who have been rumored to be entering the MOCVD market for many years.  That said, as device sizes continue to shrink, production techniques tend to move more toward semiconductor processes and that gives semi tool vendors an entry point into the equipment supply chain as it morphs into a semiconductor-like market in 2025 and beyond.  We doubt the big semiconductor equipment players will miss an opportunity to capture some of that potential business.

3D TVs have come and gone and VR, well if you don’t mind wearing a set of goggles, it can be effective for those who are not affected by motion sickness and visual disorientation, but there is another variety of 3D that is gaining popularity and does not need glasses, helmets, or anything unusual to view, although you will certainly not be sitting in your living room when watching it.  This form of 3D is becoming popular in China, although certainly not limited geographically, but at the moment its not for home use but rather for advertising, and is very effective at attracting attention.
The technology, loosely called naked eye 3D is based on a combination of hardware and software that is used to create short video advertisements on specialized massive LED displays found in public locations.  The LED displays themselves are huge, usually covering the face and side of a large building, a story or so above the street, but tend to be far denser than the usual ‘Time Square’- like advertising boards in terms of the number of LEDs.  Of course this makes for complex installations and very large numbers of smaller LEDs, making the projects expensive, but the advertising potential for these outdoor displays is quite high.
The technology is not just the hardware, as the image creation itself must be carefully controlled, but the techniques for creating naked eye 3D content are not a complex as one might think.   The most typical way to create depth in a flat image is by the use of occlusion relationships, such as superimposing lines in front and behind two identical images as shown in Fig. 2.  While nothing in the image has actually changed, our brain takes the lines as cues and we perceive the right image as being closer.  Unfortunately superimposing lines across video content is not an effective method for creating 3D content, but there are other ways to produce 3D on a 2D plane.

Bezels, the outer, non-display part of many displays, are considered bad by most consumers.  Everybody wants bezel-less smartphones and TVs that make images look like they are floating in space, but bezels can be used to create 3D images in a 2D world by creating ‘virtual bezels’.  By creating a ‘virtual bezel’, essentially a black frame around a video image, our brains can again be fooled into thinking that images are ‘coming out of the screen’.  Here’s how that works:

Here’ is our tree again in a normal screen view, but if we ‘expand’ that black border, not physically, but as part of the video, we can create the same effect as superimposing lines might have.  Note that using the term bezel might be confusing here, as this is really a visual ‘bezel’ created during editing rather than a physical one, but you can see that by placing objects behind or in front of the ‘virtual bezel’, they can be made to look like they are in front of the display, rather than on a flat plane.

​There are other motion parallax techniques that can be used to enhance the realism of moving images, but as the pixel pitch (distance between each LED pixel) decreases in such displays, the realism increases and by bending the display around the corner of a building, the effect of depth is further increased.  Again these displays are all 2D in that they are flat in the same way a TV screen is flat, but the image does not look flat.  Most of the videos use less than the full screen in order to capitalize on the ‘virtual bezel’, which certainly works best at night, but we have seen compositions where the virtual bezel looked exactly like the side of the building, which made the imagery very effective during the day.  More to come.
Watch the short videos below to see this very unusual effect.
https://youtu.be/rScf4OocI7E
https://youtu.be/nWKtIGVPbyc
https://youtu.be/Gzw5240T6P8
 
 

​Samsung Electronics (005930.KS) announced preliminary 3Q guidance of 73t won ($61b US), up 14.7% q/q and 9.0% y/y/, with operation profit of 15.8t won ($13.2b US), up 25.7% q/q and up 27.9% y/y.  Expectations for sales were between 73.5t won and 74.6t won, with operating profit expectations between 15.75t won and 16.1t won, which puts results in-line to slightly below consensus.  While the company does not give details other than the above until the full release later this month, expectations are that strength came from the semiconductor segment, particularly DRAM, with currency on the positive side.  While 3Q numbers were strong expectations are that it will be difficult for Samsung to maintain as much momentum in semiconductors as has been seen recently as CE demand has been weakening. 
On a segment basis, expectations are that Samsung’s semiconductor segment generated between 9.7t won and 10t won of the 15.8t won operating profit, up 3t won from 2Q, and that the IM segment (IT/Mobile) generated 27t to 28t won in sales, up 4t to 5t won but ~3.6t won in operating profit, which assumes sequentially lower segment margins.  Aggressive marketing for the foldable segment is expected to be the reason for the decline.
The stock itself peaked on the first day of the year and is down 19.48% to date despite the results this year and dropped slightly on today’s announcement, with sentiment leaning toward “the best is over”, the same sentiment that has been pressuring the stock for most of the year.  Samsung has been facing problems with its mobile division, and while foldable devices are doing well, Samsung has been losing mobile share to Chinese brands.  With global TV set sales slowing and notebook demand weakening a bit, that negative sentiment has only increased despite the best operating profit performance in over two years, but expectations for the company’s full report are for caution concerning semis, which will likely do little to remove the cloud over the stock.  

​In May we indicated that Intel (INTC) had approached one of its packaging suppliers, Samsung Electro-Mechanics (009150.KS) and was in discussions about investing what is expected to be hundreds of millions of dollars in FC-BGA, a connection technology used in high pin count devices such as server and PC CPUs.  There are only a few FC-BGA suppliers of note, which limits Intel’s potential choices, with some assuming the company has already had discussions with its two larger FC-BGA suppliers Japan’s Ibiden (4062.JP) and Shinko Electric (6967.JP).  While Samsung EM has been a smaller supplier to Intel than Ibeden and Shinko, there are other smaller players that could potentially expand capacity over time, such as Daeduck Electronics (353200.KS) and LG Innotek (011070.KS) in Korea, Nanya (2408.TT) and Unimicron (3037.TT) in Taiwan, and AT&S (ATS.AV) in Austria, especially as Intel would compete against Samsung EM’s largest customer, parent Samsung Electronics.
While unconfirmed by Intel, the big investment in FC-BGA that we noted (10-01-21) that Samsung EM has recently decided to make, “dedicating some or all of these lines to a US Chip Company” would seem to indicate that at least some agreement has been made, although no timetable for the expansion was indicated.  But it seems that the activity in the FC-BGA space continues as LG Innotek has indicated that they have started investing in the segment, more specifically for automotive components, which could make a bit easier to compete against Samsung EM, who is more focused on more typical CE products and servers.   The company was originally expected to spend between 400b and 500b won ($334m - $418m US) for the project but parent LG Group (pvt) thought that might not be enough to compete with Samsung EM and has now committed 1.1t won ($919.6m US) to the FC-BGA project, and some of the smaller South Korean producers have also indicated increased spending in the space.
With new capacity coming on line over the next few years, one would expect these companies to be able to overcome any shortages that might occur however as the boards become more complex (multi-layered) yields decline and unit volumes drop, which is the rationalization for the new capacity.  Whether that works as smoothly as expected remains to be seen, but during periods of component shortages such as the PCB industry is currently facing, boards are more willing to roll the dice.

September saw an increase in almost all of the 5G indicators that we track.  Overall device offerings increased 7.0% m/m and 138.7% y/y, with a 2 year increase of 721.7%, while 5G smartphones increased 9.4% m/m and 161.5% y/y (1175.6% over two years).   5G CPE devices increased 6.2% m/m and 59.3% y/y and routers, laptops and tablets all increased m/m.  Only 5G in-vehicle routers remained flat m/m, but that is a product that tends to be based on vehicle model releases.
Despite the rather weak smartphone market, 5G smartphone offerings continue to grow, and in China, where even against easy comparisons, smartphone shipments are only up 12.3%, 5G smartphones made up 47.4% of new models in August and will likely reach over 50% into the holiday season.  In terms of 5G itself, the ratio between sub6 and mmWave phones remains consistent at ~55%/45%, with sub6 the leader, which is surprising given the far more extensive buildout of sub6 5G capacity.  We expect that the cost of dual modem and antenna systems has dropped to the point where the decision to offer both is the logical choice to satisfy as many customers as possible.
 

Along with the iPhone 13 on September 14, Apple (AAPL) also announced the latest version of the iPad Mini, the iPad Mini 6, an 8.3” tablet designed to be a handheld device a bit larger than a premium smartphone but less than a full size tablet.  Apple made a number of improvements in the Mini 6, particularly with the chipsets, but along with those improvements came the inevitable bugs that are a part of almost any hardware upgrade, and it doesn’t take long for users/bloggers to dig them out once the device has been delivered (September 24).
The most prominent of the bugs is called ‘jelly scrolling’, aka ‘scan out skewing’.  This phenomena occurs when a user is scrolling the screen in portrait mode, with one side of the display refreshing more slowly than the other, making images and text jiggle.  It doesn’t happen as noticeably when the device is scrolled in portrait mode.  This issue is actually one that occurs in many devices but is almost invisible, however due to a change Apple made in the internal design of the IPad Mini 6, it has become an issue.  In a typical device, such as a smartphone or a monitor, the controller board is oriented in the same direction as the way the display is typically scrolled, meaning as you scroll down, the screen is being refreshed line-by-line vertically.  This ‘hides’ much of the action of the repainting of the screen as you move the text or image.  In the iPad Mini 6 however, Apple changed the orientation of the controller board, which causes the screen to be refreshed horizontally.  When you scroll up or down in portrait mode the controller is repainting the screen horizontally, rather than following your scroll, and the text or image jiggles as you scroll.
Apple has responded to complaints by saying that this is a ‘normal’ function of how screens refresh, and to a degree they are right, as the iPad Pro also has the controller board mounted the same way, but the faster refresh rate of the higher cost iPad Pro makes the jelly scroll unnoticeable.  Some have speculated that Apple might have used a less expensive, lower spec display, which might be exaggerating the issue, but that remains unconfirmed.   Since a fix would seemingly require a redesign of the iPad Mini 6 internals, we expect there is little users or Apple can do about the problem, or than to use the Mini 6 in landscape mode whenever possible.
Within the last few days however, a new issue has popped up concerning the Mini 6. There have been reports that screen discoloration and distortion can occur when the user uses extra touch pressure, such as when you press and hold.  When the user reporting the issue received a replacement, the same problem occurred, although Apple was willing to make that replacement for this problem, while not for the jelly scrolling issue.  As this issue does not seem to be widespread, at least currently, we doubt Apple will make any changes to its display suppliers, but such potential defects do have an effect on Apple’s supply chain once they determine from where the problem occurs.  If it is a touch issue or a display issue it could have an effect on Apple’s supplier choices going forward, but with only a few days of dealing with the problems, we expect Apple is trying to determine whether the new problem has a single component root and if so is it limited to a small number of units or the entire line.
https://youtu.be/WWKb_khoFzw - Jelly Scrolling - First 20 seconds (Ctl-Click)